Load-Aware Radio Access Selection in Heterogeneous Terrestrial Wireless Networks
-
Upload
aircc-ijcnc -
Category
Documents
-
view
221 -
download
0
Transcript of Load-Aware Radio Access Selection in Heterogeneous Terrestrial Wireless Networks
8/3/2019 Load-Aware Radio Access Selection in Heterogeneous Terrestrial Wireless Networks
http://slidepdf.com/reader/full/load-aware-radio-access-selection-in-heterogeneous-terrestrial-wireless-networks 1/20
International Journal of Computer Networks & Communications (IJCNC) Vol.3, No.6, November 2011
DOI : 10.5121/ijcnc.2011.3606 95
LOAD- AWARE RADIO ACCESS SELECTION IN
HETEROGENEOUS TERRESTRIAL WIRELESS
NETWORKS
M. Ali, P Pillai, Y.F.Hu and Kai Xu
School of Engineering Design and Technology
University of Bradford, UK
{m.ali28, p.pillai, y.f.hu, k.j.xu }@bradford.ac.uk
A BSTRACT
Terrestrial wireless network technologies such as UMTS, WiMax and WLAN are used to provide network access for both voice and data services. In big cities the densely populated areas like town centres,
shopping centres and train stations which have coverage of multiple wireless networks, a large number
of mobile users may be connected to the more common UMTS, thereby creating an unbalanced load
across these wireless networks. This high load variation can be balanced by moving mobile users from
heavily loaded networks to least loaded networks which involves execution of vertical handovers.
Seamless vertical handovers across different wireless networks may be achieved using the IEEE 802.21
Media Independent Handover (MIH) specifications. Radio Access Technology (RAT) selection
techniques aim to find the most suitable network that a mobile node should be connected to, for
achieving seamless services and meeting the QoS requirements of the user. Traditional RAT selection
algorithms are mainly based on the Always Best Connected (ABC) paradigm whereby the mobile nodes
are always directed towards the available network which has the strongest and fastest link. This however
could create a high variation among the load across the different co-located networks; which cause
congestion on overloaded network and eventually increase the call blocking and call dropping probabilities. The unbalanced load situation in co-located networks also causes the poor radio resource
utilization as some networks remain under loaded and some become over loaded. There is a need for the
load balancing strategies to efficiently utilize the available radio resources and avoid the unwanted
congestion situations on overloaded wireless networks. This paper proposes a novel network load aware
RAT selection technique in heterogeneous terrestrial wireless networks to minimize the load variation in
heterogeneous networks to a suitable level and describes how the MIH framework may be extended to
make seamless vertical handover load conscious. The proposed strategy for load balancing consists of
two different algorithms; first located in mobile user and the other at network entity such as RNC, BS or
AP. The proposed method considers the network type, signal strength, data rate and network load as
primary decision parameters for RAT selection process and tries to maintain the load equilibrium on all
networks which have common or overlapped coverage areas. Different attributes like load distribution in
all wireless networks, average end-to-end delays, jitters and average handover latencies have been
observed to evaluate the effects of load balancing in considered scenarios.
K EYWORDS
Load balancing, radio resource management, heterogeneous wireless networks, load balancing in
wireless networks, vertical handovers, terrestrial wireless networks load balancing.
8/3/2019 Load-Aware Radio Access Selection in Heterogeneous Terrestrial Wireless Networks
http://slidepdf.com/reader/full/load-aware-radio-access-selection-in-heterogeneous-terrestrial-wireless-networks 2/20
International Journal of Computer Networks & Communications (IJCNC) Vol.3, No.6, November 2011
96
1. INTRODUCTION
Modern mobile devices like cell phones, PDA’s, Tablet PCs already support multiple wireless
technologies like UMTS, WLAN and Bluetooth and in the very near future would also support
WiMax. While most of these devices are able to scan the different available networks the userwould manually select which network he or she may want to use. It is envisaged that in the nearfuture these devices may be able to apply some complex Radio Access Technology (RAT)
selection techniques to find the most suitable network. Such a RAT selection technique may
need to consider various parameters like received signal strengths, errors rates, costs, user
preferences, QoS requirements, etc. Such a RAT selection technique would not only play an
important part when a user switches on his or her mobile device but also when the user movesaround. While most of the current day mobile networks already support seamless handovers,
these are restricted to handovers within the same technology, i.e. horizontal handovers. It is
envisaged that to efficiently use the network services the future mobile devices shall also
support handovers across different radio access technologies. This process of switching mobile
devices connectivity from one technology to another type of technology is called vertical
handover.The joint call admission control (JCAC) algorithm for future heterogeneous wirelessnetworks is envisioned as user-centric. User centricity implies that user’s preferences are
considered in decision making for RAT selection. However user-centric JCAC algorithms often
lead to highly unbalanced networks load, which cause congestion on overloaded network and
eventually increase the call blocking and call dropping probabilities. The unbalanced load
situation in co-located networks also causes the poor radio resource utilization as some
networks remain under loaded and some get over loaded.The load balancing strategies are
required to efficiently utilize the available radio resources and avoid the unwanted congestion
situationsdue to overloaded wireless networks.
Proposed solution for load balancing involves the utilization of IEEE 802.21 Media
Independent handover (MIH) [1] for moving load (mobile nodes) between different wireless
networks.The MIH framework defines a common interface between different link layertechnologies for the support of seamless mobility between heterogeneous IEEE-802 networks
and between IEEE-802 and other mobile wireless technologies. This unified interface is
presented as an abstraction layer function, the Media Independent Handover Function (MIHF),
for handover detection, initiation and decision via Layer 2 triggers. The MIH provides the
seamless mobility to mobile nodes between heterogeneous networks using a set of services
known as Media Independent Command Service (MICS), Media Independent Event Service
(MIES) and Media Independent Information Service (MIIS). Brief description of IEEE 802.21
is given in the section 3.
This paper presents a novel RAT selection technique which uniformly distributes the network
load between co-located heterogeneous wireless networks. It utilizes MIH to seamlessly
handover mobile users between heterogeneous wireless networks for load balancing purpose.
The advantage of this approach is that it minimizes the call blocking and dropping probabilities,
number of packet drop/lost anddelays during the handover process and enhances the network
utilization by continuously balancing the load in co-located networks. The proposed load
balancing approach monitors and controls the network load from both side (mobile node and
network side).The rest of the paper is organised as follows; Section 2 presents an overview on
handovers in wireless networks and IEEE 802.21 MIH specification, Section 3 briefly describes
some existing load balancing techniques that are proposed for wireless networks. The target
8/3/2019 Load-Aware Radio Access Selection in Heterogeneous Terrestrial Wireless Networks
http://slidepdf.com/reader/full/load-aware-radio-access-selection-in-heterogeneous-terrestrial-wireless-networks 3/20
International Journal of Computer Networks & Communications (IJCNC) Vol.3, No.6, November 2011
97
network architecture and proposed load balancing algorithms are explained in detail in Section
4. Section 5 presents the various simulation scenarios and obtained performance evaluationresults and finally the conclusion is presented in the Section 6.
2. HANDOVERSIN HETEROGENEOUS NETWORKS
2.1 Handover terminology
The handover procedure can be divided into three phases namely handover initiation, handover
decision and handover execution. The first two phases (handover initiation and decision) can be
comprised of any four basic schemes which are: mobile controlled handover (MCHO), network controlled handover (NCHO), mobile assisted handover (MAHO) and network assisted
handover (NAHO). There can be other hybrid schemes evolved from these basic schemes such
as mobile assisted network controlled and network assisted mobile controlled handovers. The
handover execution phase follows the initiation and decision phases. In execution phase the
mobile node establishes handover connections with target network and releases all connectionswith serving network, which requires signalling exchange procedure between mobile node and
the network entity (target or serving network). The signalling exchange procedure between themobile node and network for handover execution can be of two types such as backward and
forward. The backward scheme utilizes serving network link for signalling exchange, whereas
the forward scheme establishes and uses new signalling link with target network [2, 3, 4].
Handover can also be classified in to three categories namely, hard handover, soft handover and
softer handover. In hard handover a mobile device disconnect itself from the current servingnetwork before connecting to the target network leading to the break-before-make handover
scenario. In contrast, soft handover is a make-before-break handover where a mobile device
connects to target network before disconnecting itself from current serving network. In case of
softer handover, the mobile device stays connected to the serving network but retunes its
communication frequency or communication channel. Softer handover addresses macro-level
mobility. Handover can also be categorized as horizontal and vertical handovers, in whichhorizontal handover represents the process of migrating mobile device from one network to
another provided that both serving and target networks are of the same type, whereas in vertical
handover the target and serving networks are of different types.
2.2 IEEE Media Independent Handover (MIH)
The IEEE802.21 Media Independent Handover (MIH) standard defines a common interface
between different link layer technologies such as IEEE 802.11 [5, 6], IEEE 802.16 [6, 7] and
mobile cellular technologies like UMTS [8-11] to support seamless mobility between them.This common interface is provided as an abstraction layer called Media Independent Handover
Function (MIHF). The MIHF receives media independent commands from higher layers and
translate them to media dependent commands for the link layer and similarly receives eventsfrom different link layer technologies and maps them onto corresponding media independent
events. Figure 1 shows the IEEE 802.21 reference model with its associated Service Access
Points (SAPs). MIH provides a set of services to facilitate the media independent handoverprocess, namely, the MIH Event Service (MIES), the MIH Command Service (MICS) and the
MIH Information Service (MIIS). The MIES reports events on dynamic changes in link
characteristics such as link status and link quality to upper layers through the MIHF. The MICS
is used to collect information about the status of connected links. Upon receiving event
8/3/2019 Load-Aware Radio Access Selection in Heterogeneous Terrestrial Wireless Networks
http://slidepdf.com/reader/full/load-aware-radio-access-selection-in-heterogeneous-terrestrial-wireless-networks 4/20
International Journal of Computer Networks & Communications (IJCNC) Vol.3, No.6, November 2011
98
notification, an MIH user uses the MICS to pass link commands to the lower layers through the
MIHF to manage and control the link layer behaviour for handover decision. The MIISprovides capability for obtaining the necessary information for handovers, which includes
neighbouring networks and link layer information. The information gathered using MIIS is
used to assist network discovery and selection which enables efficient decision making forhandover. The MIH user shown in the Figure 1 uses the services provided by the MIHF.
Figure 1: MIH Reference Model [1]
Each SAP consists of a set of service primitives that specify the interactions between the MIHF
and other entities in the MIH reference model. All exchanges of MIH messages between theMIHF and other functional entities occur through service primitives, which are grouped in
SAPs. Three SAPs are included in the MIH reference model:
• MIH SAP acts as a media independent interface between the MIHF and higher layers
of protocol stack for mobility management. The higher layers after subscribing with the
MIHF as MIH users can send link commands and receive link events through the
MIHF from a particular link using service primitives of the MIH SAP.
• MIH Link SAPis a generic name for the media-dependent interface between the MIHF
and the lower layers of the local node protocol stack. The MIH Link SAP maps on
different link specific SAPs for each link layer technology. For 802.11 it maps onto the
MLME_SAP, for 802.16 on the M_SAP and for UMTS on the MIH_3G_LINK_SAP.
• MIH NET SAP is a generic name for the media-dependent interface between the
MIHF and the transport elements, which supports the exchange of MIH information
and messages with remote MIH Function instances.
3. LOAD BALANCING TECHNIQUES
Usually more than one wireless networks may provide coverage to any given location in an
urban area. For example, when in an office, the mobile device may be in the coverage of a
UMTS mobile network and a WLAN office network. A user may manually configure to use the
UMTs network for voice services but the WLAN access for data services. If such was in the
coverage of other technologies like WiMax also, then the networks loads would not be
balanced. In such overlapping areas, a RAT selection technique is required to find the most
suitable network based on received signal strengths, errors rates, costs, user preferences, QoS
8/3/2019 Load-Aware Radio Access Selection in Heterogeneous Terrestrial Wireless Networks
http://slidepdf.com/reader/full/load-aware-radio-access-selection-in-heterogeneous-terrestrial-wireless-networks 5/20
International Journal of Computer Networks & Communications (IJCNC) Vol.3, No.6, November 2011
99
requirements, etc. However it is important that a load balancing approach is required for RAT
selection to avoid such unwanted load variation and congestion situations. This would aim tobalance the load across the different networks, such that we do not have a situation where one
of the networks is over utilised, while the other are underutilised. The process of load balancing
is being carried out from more than two decades in the field computing but is relatively new inwireless communication networks. In computing, load balancing techniques are used
extensively for balancing the load across different back-end servers. Some approaches for
balancing the load in both homogeneous and heterogeneous wireless networks have been
presented in literature. While some of these consider mobile users based load balancing, other
consider network based RAT selection
The load balancing approaches presented in [12] and [13] have considered load balancing in
homogenous network targeting WLAN. The approach in [12] considers the received signal
strength indicator (RSSI) value to distribute the load between different access points (AP’s)
which have overlapping coverage areas.This approach uses the two values in balancing the load
which are RSSI between mobile station (MS) and AP and the average RSSI value of all the
MS’s currently connected with AP. The method given in [13] considers both RSSI and thenumber of MS associated with AP which makes it much effective for load balancing.The
technique used in [14] presented a solution for load balancing in homogeneous wireless
networks, by utilizing genetic algorithm. As the genetic algorithm’s convergence directly
proportional to the size of population (mobile nodes and APs) therefore this approach is
effective for WLAN networks and not for the heterogeneous wireless environment where
population size iscomparatively large due to large coverage areas. All approaches given in [12,
13, 14] were designed to enhance the performance for homogeneous network environment
particularly WLAN.
In [15] load balancing approach has been presented which targets the proxy mobile ipv6
(PMIPV6) domain using MIH for heterogeneous networks. A comparison has been made
between the scenario performing load balancing in extended PMIPV6 for handover signallingand the scenario using MIH signalling for load balancing. It was shown in the results that use of
load balancing improves the efficiency whereas, MIH based load balancing improves data rate
as compared to extended MIPV6 based load balancing. This approach has two disadvantages
when considering load-aware RAT selection: i) it is specifically designed for a MIPV6
architecture using Local Mobility Agent (LMA) and a new entity called Mobile Access
Gateway (MAG) in the network. ii) This approach triggers load balancing on two conditions,
first when a particular network load exceeds a particular threshold and secondly when whole
system load variation exceeds a particular threshold limit; which triggers load balancing after
the network becomes overloaded.In [16] a general set of algorithms have been proposed which
considers battery power of mobile users, received signal strength and load on available points
of attachments in handover process to balance the load in co-located networks overlapping their
coverage areas. Simulation results have been shown for WiMax and WLAN with differentscenarios using proposed algorithms.While this approachalso utilizes MIH like our proposed
approach in this paper, in their approach load balancing is done only at network side such as
BS, AP or RNC without any interaction with the mobile node. On the other hand ourproposed
approach considers both mobile nodes and network entities such as AP, BS and RNC for load
balancingthereby resulting in more efficient load balancing acrossthe neighbouring networks.
In [17] a detailed algorithm has been presented for network selection in heterogeneous wireless
networks. The algorithm presented in [17] has been divided into two parts, one runs at mobile
8/3/2019 Load-Aware Radio Access Selection in Heterogeneous Terrestrial Wireless Networks
http://slidepdf.com/reader/full/load-aware-radio-access-selection-in-heterogeneous-terrestrial-wireless-networks 6/20
International Journal of Computer Networks & Communications (IJCNC) Vol.3, No.6, November 2011
100
terminals and other part of algorithm runs at network entity such as basestation (BS) or access
point (AP). This approach considers received signal-strength, battery power, speed, andlocation of mobile user but does not considers MIH which could have improved the handover
process while moving the mobile nodes between different networks. In [18] a next generation
networks (NGN) based approach has been presented in which hierarchical joint call admissioncontrol algorithm is extended to send newly added load reports from hierarchical call admission
control (HCAC) entity to vertical call admission control entity (VCAC). The main goals of
proposed approach in [18] are simplicity and scalability, however this approach performs
balancing of load periodically and therefore may not performs very efficiently with abrupt load
changes in different sub networks in the hierarchy. In [19] a Markov chain based model for
load balancing and QoS based CAC has been presented and comparisons have been made
between the results of load balancing based CAC and QoS based CAC algorithms. It has been
shown clearly in results that load balancing based CAC performs better as compared to the QoS
based CAC in throughput, call blocking and call dropping probability graphs as the call arrival
rate increases. The load balancing approach presented in this paper more efficient than load
based CAC approach presented in [19] as our approach uses MIH to minimize the handover
delays when moving the mobile nodes for load balancing purpose and tends to uniformlydistribute the load among available heterogeneous wireless networks.
4. THE LOAD-AWARE RAT SELECTION FRAMEWORK
4.1 Network Architecture
Figure 2 presents the target network architecture which is considered in this paper. It shows anMIH enabled multi interface mobile device which can use any of the three available wireless
terrestrial networks supported by its interfaces. The access network of each technology such as
UMTS, WiMax and Wi-Fi is connected to internet. There is also a correspondent node located
behind the internet as shown in the Figure 2. The mobile node can communicate with the
correspondent node over the internet using any available network which is supported by itsinterfaces.The mobile node handovers to different available networks while moving from
coverage area of one network to another and during this mobility it can maintain the
communication with correspondent node.
8/3/2019 Load-Aware Radio Access Selection in Heterogeneous Terrestrial Wireless Networks
http://slidepdf.com/reader/full/load-aware-radio-access-selection-in-heterogeneous-terrestrial-wireless-networks 7/20
International Journal of Computer Networks & Communications (IJCNC) Vol.3, No.6, November 2011
101
Figure 2: Target network architecture
The load balancing algorithms are located at the MIH user in MIH reference model as
represented by the Figure 2. MIH user is selected for the load balancing process origin as MIH
user is the central control point for triggering and handling MIH signalling as described in the
previous section and in [1]. In mobile user the load balancing algorithm shown in Figure 3 is
adopted and in RNC/BS/AP the load balancing algorithm for the network entity shown in
Figure 4 is adopted.
4.2 Load balancing algorithms
This section describes the proposed load-aware RAT selection algorithm.The proposed
algorithm considersthe network type, signal strength,data rate and network load as primary
decision parametersfor RAT selection process andtries to maintain the load equilibrium on allnetworks which have common or overlapped coverage areas. It is assumed that all considered
networks and mobile nodes support the IEEE 802.21 MIH. The IEEE 802.21 MIH standard has
been brought into use for seamless vertical handover operationsof mobile nodes between the
co-located wireless networks.The proposed approach has taken advantage of MIH media
independent information service (MIIS)specifically for exchange of network load information
besidesexchanging other network related information like link type,link data rate,link
capability, offered security and QoS andcost[1].
The proposed RAT selection framework consists of two load aware algorithms, one running on
mobile device and other running on network entity like Radio Network Controller (RNC),
WiMax BS or WLAN AP. The flow chart shown in Figure 3 represents the proposed
algorithm’s which runs at mobile device. At the mobile device, the proposed technique firstmakes a list of available network IDs which are visible to mobile device such that received
signal strength from those networks is higher than the minimum threshold. In next step load
value of each network in the list is obtained from MIIS and compared. Then in following step it
compares the data rate offered by each network in the list. The most preferred network from the
list is the one with lowest load and highest offered data rate. The second algorithm shown in
Figure 4 runs in network side.
8/3/2019 Load-Aware Radio Access Selection in Heterogeneous Terrestrial Wireless Networks
http://slidepdf.com/reader/full/load-aware-radio-access-selection-in-heterogeneous-terrestrial-wireless-networks 8/20
International Journal of Computer Networks & Communications (IJCNC) Vol.3, No.6, November 2011
102
Figure 3: Load balancing algorithm in the mobile device
In the network entity like RNC, BS or AP the load balancing algorithm continuously keeps on
updating the MIIS about its current load status and receives load information of its
neighbouring networks. This updating process runs on every time when a new connection starts
or ends in the network. The most loaded network entity start moving out the suitable mobile
users to appropriate networks, if the load variation is gone higher than threshold of 50% free
resources margin, such that the percentage of free resources in one network is greater than or
equal to the double of available resources percentage at any other network. Load balancing
algorithm keeps on migrating out the suitable mobile nodes from over loaded network to the
least loaded networks until the load in over loaded network becomes equal to or lesser than the
average load in all the neighbouring networks of overloaded network.
8/3/2019 Load-Aware Radio Access Selection in Heterogeneous Terrestrial Wireless Networks
http://slidepdf.com/reader/full/load-aware-radio-access-selection-in-heterogeneous-terrestrial-wireless-networks 9/20
International Journal of Computer Networks & Communications (IJCNC) Vol.3, No.6, November 2011
103
Figure 4: Load balancing algorithm at the Network side (RNC/BS/AP)
4.3 SAPs primitives mapping
Table 1 represents the mapping of the SAP primitives for IEEE 802.16, IEEE 802.11 and
UMTS onto the MIH Link SAP primitives [1, 9]. These mappings are used in the handover and
simulation scenarios described in the upcoming sections. The MIH SAP primitive
Link_Detected is triggered by the link specific primitives when the network for that technologybecomes reachable. There is no corresponding primitive for Link_Detected in IEEE 802.11 and
UMTS, however in IEEE 802.16 the C_SAP primitive C-HO-RSP for handover scanning
triggers Link_Detected towards the MIHF. The Link_UP primitive of the MIH Link SAP is
triggered by C-NEM-RSP which is the C-SAP primitive in IEEE 802.16 for registration andranging response, MLME-LinkUp.indication in IEEE 802.11 and RABMSMACTIVATE in theUMTS when a mobile node gets registered with network.
Table 1: MIH primitives mapping
MIH Link_SAP
Primitives
IEEE 802.16
SAP
IEEE 802.11 SAP UMTS
SAP
Link_Detected C-HO-RSP(HO-
Scan)
N/A N/A
Link_UP C-NEM-RSP
(Registration)
MLME-LinkUp.indication RABMSM-
ACTIVATE
Link_Down C-NEM-RSP
(Deregistration)
MLME-
LinkDown.indication
RABMAS-
RAB-
RELEASE
Link_Going_Down N/A MLME-
LinkGoingDown.indication
N/A
Link_Parameters_Report C-HO-RSP
(HO-Scan)
MLME-
MREPORT.indication
RABMSM-
MODIFY
8/3/2019 Load-Aware Radio Access Selection in Heterogeneous Terrestrial Wireless Networks
http://slidepdf.com/reader/full/load-aware-radio-access-selection-in-heterogeneous-terrestrial-wireless-networks 10/20
International Journal of Computer Networks & Communications (IJCNC) Vol.3, No.6, November 2011
104
The Link_Down is MIH link SAP primitive triggered towards MIHF by IEEE 802.16 C_SAP
primitive C-NEM-RSP with deregistration response, IEEE 802.11 primitive MLME-LinkDown.indication and UMTS primitive RABMASRAB-RELEASE-RELEASE when a
mobile node deregisters or gets disconnected from the network. The Link_Going_Down has no
corresponding primitives in IEEE 802.16 and UMTS, but in IEEE 802.11 MLME-LinkGoingDown.indication triggers the Link_Going_Down towards the MIHF when signal
strength of the serving network becomes weaker. The Link_Parameters_Report indicates
changes in link conditions. For the load balancing an additional parameter named as load with
data type integer has been introduced in information element (IE) of MIIS. The MIIS primitives
MIH_Get_Information.indication and MIH_Get_Information.confirm carry this value from
MIIS to mobile node.
4.4 Detailed handover procedures
The handover scenarios considered in this paper cover the vertical handover procedures
between UMTS, WiMax and Wi-Fi in such a way that it covers circumstances which trigger
handover when mobile node enters and leaves these networks. Figure 5 shows the procedureand the SAP primitives involves in the handover from UMTS to WiMax. The step 1 in Figure 5
informs the MIHF that the mobile node is registered with the UMTS network. Step 2 informs
the MIH User about the activation of link on mobile node’s UMTS interface. In step 3, a TCP
connection is established between the mobile node and the TCP source using the UMTS
network. Step 4 shows WiMax interface receives broadcast messages from WiMax BS. Step 5
signals the MIHF in the mobile node about WiMax network detection. Step 6 informs MIH
User about WiMax link detection. Insequence of steps from 7 to 10 the mobile node acquire the
neighbouring networks information from MIIS using a set of MIIS primitives, and step 11
checks the availability of required resources in WiMax network. Step 12 decides whether or not
to perform handover to WiMax. In steps 13 to 22, the mobile node registers itself on WiMax.
Step 23 is to establish a TCP connection between mobile node and TCP source using WiMax
network. In step 24, mobile node releases its connections from the UMTS network.
Figure 5: Handover from UMTS to WiMax
8/3/2019 Load-Aware Radio Access Selection in Heterogeneous Terrestrial Wireless Networks
http://slidepdf.com/reader/full/load-aware-radio-access-selection-in-heterogeneous-terrestrial-wireless-networks 11/20
International Journal of Computer Networks & Communications (IJCNC) Vol.3, No.6, November 2011
105
Similarly, Figure 6 shows the SAP primitives used in the handover procedure from WiMax to
Wi-Fi. As represented by Figure 6 the 802.11 MAC layer in the mobile node, after detectingand registering with Wi-Fi network, it sends MLME-LinkUp.indication message to the MIHF.
In step 3, MIHF sends MIH_Link_UP.indication to MIH User. A set of messages from step 4 to
7 acquire the neighbouring networks information. Step 8 checks for the required resources inWi-Fi for handover. The MIH User decides whether to perform handover or not in step 9. Steps
10 to 12 show the handing over of the connections to the Wi-Fi network. Finally, step 13 and
step 14 make sure that traffic flow has been re-establish between mobile node and TCP source
and then release bindings with UMTS network. Figure 7 shows the handover procedure when
mobile user moves away from the Wi-Fi coverage area and enters a WiMax coverage area.
The first step in Figure 7 is the message MLME_MREPORT.indication from MAC Wi-Fi to
MIHF. This is the periodic message which carries parameters of link. In step 2 the MIH User is
being updated with link parameters report. Step 3 shows the message link-Going_down from
Wi-Fi MAC to MIHF, which represents that mobile node, is gradually losing the connectivity
with Wi-Fi. Step 4 informs the MIH User about link going down event. From step 5 to step 8
are used to acquire neighbouring networks information from MIIS. Step 9 shown as bubblerepresents the process of scanning on all interfaces supported by mobile node. Step 10 and step
11 are for selecting the WiMax network and handover all connections to WiMax.
Figure 6: Handover from WiMax to Wi-Fi
The last message sequence chart shown in the Figure 8 represents the handover procedure whenmobile user moves from a WiMax-UMTS common coverage area to an area covered by UMTS
only. Here mobile node handovers to the UMTS from the WiMax network. Step 1 represents
that the mobile node has lost the WiMax connectivity. In step 2 the WiMax MAC sends link
down equivalent primitive to MIHF, this triggers the MIH_Link_Down primitive from the
MIHF to MIH User in step 3. In messages from step 4 to step 6 scanning for the other available
links performed. Sequence of steps from step 7 to step 10 shows that mobile node acquire
8/3/2019 Load-Aware Radio Access Selection in Heterogeneous Terrestrial Wireless Networks
http://slidepdf.com/reader/full/load-aware-radio-access-selection-in-heterogeneous-terrestrial-wireless-networks 12/20
International Journal of Computer Networks & Communications (IJCNC) Vol.3, No.6, November 2011
106
neighbouring networks information from MIIS using MIIS primitives. Step 11 shows the
decision making process for selecting the candidate network for handover, UMTS is selected as
it is the only available network. Step 12 represents the mobile node’s handover to UMTS. In
step 13 all bindings with WiMax are being released.
Figure 7: Handover from Wi-Fi to WiMax
Figure 8: Handover from WiMax to UMTS
5. SIMULATION ARCHITECTURE&RESULTS
5.1 Simulation architecture
Figure 9 presents the simulation topology considered in this paper. Purpose for considering
particular topology for simulation is to observe the effects of load balancing in most ideal
8/3/2019 Load-Aware Radio Access Selection in Heterogeneous Terrestrial Wireless Networks
http://slidepdf.com/reader/full/load-aware-radio-access-selection-in-heterogeneous-terrestrial-wireless-networks 13/20
International Journal of Computer Networks & Communications (IJCNC) Vol.3, No.6, November 2011
107
scenarios where mobile nodes can see maximum overlapped coverage areas from different
networks. Each mobile user maintains a TCP connection with the TCP source shown in Figure9 throughout the simulation such that effects of handovers on active connections can be
measured.
The set of scenarios considered in this paper consist of a group of mobile users which travel
across the coverage areas of all three networks such as UMTS, Wi-Fi and WiMax as shown in
Figure 9. There are five scenarios simulated with different number of mobile users in group
such as 5, 10, 15, 20 and 25. All mobile users support multiple interfaces such as UMTS,
WiMax and Wi-Fi and use these interface throughout the simulation. Each of these five
scenarios simulated with load balancing and without load balancing algorithm. When using the
load balancing algorithm in RAT selection; the mobile users are assigned to the network by
considering the overall load variation in the co-located networks; therefore it maintains the load
equilibrium in the co-located networks. On other hand where load balancing approach is not
utilized the mobile users move to the network with the strong signalling strength which creates
a noticeable amount of load variation in the co-located neighbouring networks.
Figure 9: Network topology for simulation
Table 2presents the simulation parameters for the different scenarios. In simulation a group of mobile users starts from the UMTS coverage area and move together towards the WiMax
coverage area. At time 10 seconds all mobile users enter in WiMax coverage area and at
approximately at 340 seconds they leave WiMax coverage area. The Wi-Fi coverage area is
overlapped by WiMax therefore at time 60 seconds group of mobile users enters the Wi-Ficoverage area and at approximately 73 seconds all mobile users leave Wi-Fi coverage. UMTS
coverage is available to the mobile users throughout the simulation time from time 0 seconds to400 seconds. The TCP source shown in the Figure 9 maintains a TCP connection with each
mobile node throughout the simulation.
8/3/2019 Load-Aware Radio Access Selection in Heterogeneous Terrestrial Wireless Networks
http://slidepdf.com/reader/full/load-aware-radio-access-selection-in-heterogeneous-terrestrial-wireless-networks 14/20
International Journal of Computer Networks & Communications (IJCNC) Vol.3, No.6, November 2011
108
Table 2: Simulation parameters
Parameters Value
UMTS Coverage 2000m
WiMax Coverage 1000mWi-Fi Coverage 100m
Speed of mobile nodes 3 m/s
Number of connections per mobile node 1
Connection Type TCP
Total number of scenarios 5
Number of mobile users in each scenario 5, 10, 15, 20, 25
Types of technologies supported by each mobile node UMTS, WiMax and Wi-Fi
Simulation time 400 Sec.
5.2 Simulation Results
5.2.1 Load balancing performance
All scenarios discussed in the previous section have been simulated using NS2[20]. Figure 10
to Figure 13 represent the load distribution at each network in 5, 10, 15, 20 and 25 nodes
scenarios with load balancing and without load balancing.
Figure 10:Scenario 1- 5 nodes
Figure 11:Scenario 2- 10 nodes
Figure 12:Scenario 4- 15 nodes
Figure 13: Scenario 4 - 20 nodes
In Figure 10 the blue and the green linesrepresent the UMTS network with and without load
balancing.The graph shows that with load balancing UMTS network carries some load
throughout the simulation and does not stay idle at all, whereas without load balancing UMTS
network carries maximum load at the beginning and end of simulation and stays idle for a long
period of time in between; which is poor utilization of radio resource available in the form of
8/3/2019 Load-Aware Radio Access Selection in Heterogeneous Terrestrial Wireless Networks
http://slidepdf.com/reader/full/load-aware-radio-access-selection-in-heterogeneous-terrestrial-wireless-networks 15/20
International Journal of Computer Networks & Communications (IJCNC) Vol.3, No.6, November 2011
109
UMTS network. Similarly pink and yellow lines showthe network load for WiMax with and
without load balancing respectively. For WiMax the graphs shows that without load balancing
all the mobile nodes move to WiMax when it is available to them. After some time when Wi-Fi
becomes available all mobile nodes move to Wi-Fi and WiMax become idle. Upon leaving the
coverage area of Wi-Fi all the mobile nodes again join the WiMax network. With loadbalancing algorithm it can be noticed that WiMax network never stay idle as long as mobile
nodes remain in the WiMax coverage area.It shows that with load balancing the WiMax
network is utilized better as it never stays idle and not all the mobile nodes connect to it any
time. The red and the cyan lines representing load at Wi-Fi show low load value with load
balancing and high load without load balancing.
Table 3 shows the comparison of load in different networks for the simulation scenario of 25
nodes with and without load balancing. Load balancing also minimizes the connection/call
blocking and dropping probabilities by sharing the load and maximizing the available networks
resources. The term LB in the following Table 3 represents Load Balancing and NLB
represents No Load Balancing or without load balancing.
Table 3: Load Table for 25 nodes scenario Approximate
Time
Available Coverage UMTS Load WiMax Load Wi-Fi Load
NLB LB NLB LB NLB LB
0-10 seconds UMTS only 25 25 0 0 0 0
10-60 seconds UMTS & WiMax 0 12 25 13 0 0
60-73 UMTS, WiMax & Wi-Fi 0 8 0 8 25 9
73-340 UMTS & WiMax 0 12 25 13 0 0
340-400 UMTS only 25 25 0 0 0 0
It is shown in the Figure 13 that without load balancing algorithm there have been a lot of load
variation such that all the mobile users acquire the best available network, which results in one
network overloaded and other nearly idle. On other hand if load balancing algorithm is used allthe available networks are equally loaded, which gives the best utilization of availabletechnologies and avoids the situation where one network gets overloaded while other having
minimum load. For example in case of 25 node scenario without load balancing from time 0 to
10 sec UMTS has all 25 mobile users at time 10 sec all 25 users handover to WiMax and
UMTS load becomes 0. Then at time 60 sec all 25 users handover to Wi-Fi and load of WiMax
and UMTS becomes 0. When all mobile users leave the Wi-Fi coverage area at time 73 sec Wi-Fi load becomes 0 and WiMax load becomes 25. On other hand with load balancing the 25
mobile users scenario shows different load distribution. Load at UMTS goes to 25 initially but
as WiMax becomes available UMTS load becomes 12 and WiMax load becomes 13 such that
each available network gets equal load. After some time when Wi-Fi becomes available to the
mobile users then load at UMTS and WiMax becomes 8 and load at Wi-Fi becomes 9. Onleaving from the Wi-Fi coverage area the mobile users from Wi-Fi network move toUMTS and
WiMax such that the load at UMTS becomes 12 and load at WiMax becomes 13.
The reason for such variations in the results for the scenarios with and without load balancing is
that without load balancing all mobile users move to the network with strong signal strength
and high data rate, leaving other available network in that area with low or no load and over
populating the best available network as shown in the list of figures from Figure 10 to Figure
13. On other hand when applying the load balancing algorithm, it controls the unbalanced
8/3/2019 Load-Aware Radio Access Selection in Heterogeneous Terrestrial Wireless Networks
http://slidepdf.com/reader/full/load-aware-radio-access-selection-in-heterogeneous-terrestrial-wireless-networks 16/20
International Journal of Computer Networks & Communications (IJCNC) Vol.3, No.6, November 2011
110
distribution of mobile nodes to the available networks. The graphs in the figures (from Figure
10 to Figure 13) are showing that load balancing maximizes the chances of availability of any
network and minimizes network overloading situation by distributing the load in the co-located
networks. These graphs also show that with load balancing idle state of network can also be
avoided in parallel with avoiding network overloading state and hence increases the availableresource utilization efficiently.
5.2.2 End-to-end delay
Figure 14 and Figure 15 show the average end-to-end delays for 5 and 25 nodes scenarios using
load balancing algorithm and without load balancing algorithm. When there is no load
balancing applied, all the mobile devices connect to the best available network and therefore
the scenario results into lowerend-to-end delays for average transmitted packet throughout the
simulation as compared to the scenario with load balancing algorithm. When load balancingalgorithm is applied each mobile user is assigned the network by considering the load at that
network. This results in a situation where different mobile users connect to different available
networks, some of them have low end-to-end delay and some have high end-to-end delay
resulting into slightly high average packet end-to-end delay. The red line in the followingFigure 14 shows the average packet end-to-end delay in 5 nodes scenario with load balancing
and blue line represents average packet end-to-end delay with no load balancing.Figure 14 and
Figure 15 show the service degradation tradeoff in terms of increasing average packet end to
end delay when using load balancing algorithm.
Figure 14: Average packet E2E delay with 5 mobile
usersFigure 15: Average Packet E2E delay with 25
mobile users
On other hand it is also shown in the Figure14 and Figure 15 that total traffic exchanged is farmore in scenario where load balancing algorithm is applied as compared to scenario where load
balancing algorithm is not applied. The reason for this increase in total number of packet
exchange is when load balancing is applied it utilizes the bandwidth available in all the
available networks such as UMTS, WiMax and Wi-Fi. On other hand when load balancing is
not applied only the network with strong signal strength and high data rate is utilized by all
mobile users.
5.2.3 Jitters
Unlike average packet end-to-end delay the value of average packet jitters for the load balanced
scenario is lower as compared to the value of average packet jitters in the scenario where load
balancing is not applied.As in case of load balanced scenario the mobile users do not frequently
change the point of attachment or target network therefore change in the end-to-end packet
8/3/2019 Load-Aware Radio Access Selection in Heterogeneous Terrestrial Wireless Networks
http://slidepdf.com/reader/full/load-aware-radio-access-selection-in-heterogeneous-terrestrial-wireless-networks 17/20
International Journal of Computer Networks & Communications (IJCNC) Vol.3, No.6, November 2011
111
delay value or jitters remains minimum, whereas without load balancing all the mobile users
moves frequently as they see new best available network which cause higher jitters. Figure 16
and Figure 17 show the average packet jitters values throughout the simulation for 5 and 25
mobile users’ scenarios with load balancing and without load balancing.
Figure 16: Average Packet Jitters for
5 mobile users
Figure 17: Average Packet jitters for
25 mobile users
5.2.4 Handover latencies
The handover latencies have been calculated for each scenario in the simulation by taking the
difference between the time of the first packet received in the target network and the time of
last packet received on previous serving network. Figure 18 to Figure 22 show the total
handover latencies observed by each mobile user in different scenarios having different number
of mobile users. While the blue bars represent handover latencies observed by each node
without load balancing algorithm, the brown bars represent handover latencies observed by
each mobile node with load balancing algorithm.
Figure 18: HO Latencies in 5 mobile users scenario
Figure 18 shows that there are total five mobile users participating in the simulation with and
without load balancing algorithm. When load balancing algorithm is applied, One mobile user
did not perform handover at all therefore its total HO latency value is ‘0’, two mobile usershave similar HO latencies (approx. 1.2 second) and one has highest HO (1.6 second) and one
has lowest HO latency value (approx.. 0.2 second). Comparing the readings shown in Figure
10 representing the load distribution of 5 mobile users scenario and Figure 18 representing HO
8/3/2019 Load-Aware Radio Access Selection in Heterogeneous Terrestrial Wireless Networks
http://slidepdf.com/reader/full/load-aware-radio-access-selection-in-heterogeneous-terrestrial-wireless-networks 18/20
International Journal of Computer Networks & Communications (IJCNC) Vol.3, No.6, November 2011
112
latencies of 5 mobile users scenario, it can be concluded that without load balancing all five
mobile users perform equal number of handovers and with load balancing at least one mobile
user stays in UMTS network throughout the simulation.Figure 10 shows the load distribution at
5 nodes scenario represents that at least five different levels of total HO latencies values are
expected which is validated in the Figure 18 showing the total handover latencies in 5 mobileusers scenario. Similar to the bar graph shown in the Figure 18, the Figure 19 to Figure 22
show total handover latencies observed by each mobile user in 10,15,20 and 25 mobile users
scenarios respectively.
Figure 19: HO Latencies for 10 mobile users
Figure 20: HO Latencies for 15 mobile users
Figure 21: HO Latencies for 20 mobile users
Figure 22: HO Latencies for 25 mobile users
As it is describedearlier in this section that different attributes like load distribution in all
wireless networks, average end-to-end delays, jitters and average handover latencies have beenobserved to evaluate the effects of load balancing in considered scenarios.It is shown in Figure
10 to Figure 13 and in Table 3 that without load balancing the networks with higher offered
data rates are preferred by the mobile users and therefore networks like WiMax and WiFi
become overloaded when group of mobile users enters their coverage areas. The networks with
lower offered data rate are given lower priority therefore in overlapped coverage areas UMTSremains under loaded; which is poor utilization of available radio resources. As congestion at
overloaded networks increases,this reduces the network availability and increases the call
dropping and blocking probability. However with load balancing the load distribution remains
uniform and all networks share load in overlapping coverage areas. The average end-to-end
delays with load balancing is slightly increased in the target scenarios as load balancingalgorithm distributes the mobile users load uniformly between networks having overlapped
coverage areas. This prevents some of the mobile users from performing unnecessary handover,
even if the target network is offering higher data rates and low delays.Average jitters value for
the scenarios with load balancing is lower as compared to the average jitters value in the
scenarios without load balancing. The reason for this is that with load balancing only
fewmobile users perform handover and rest do not change the point of attachment which does
8/3/2019 Load-Aware Radio Access Selection in Heterogeneous Terrestrial Wireless Networks
http://slidepdf.com/reader/full/load-aware-radio-access-selection-in-heterogeneous-terrestrial-wireless-networks 19/20
International Journal of Computer Networks & Communications (IJCNC) Vol.3, No.6, November 2011
113
not change the end to end delays frequently. The average handover latencies observed by each
mobile user in scenarios with load balancing is lower as compared to the scenarios without load
balancing as load balancing algorithm reduces the number of handovers for most of the mobile
users.
6. CONCLUSION
In this paper a load aware RAT selection algorithm has been presented with the comparison of
results generated by simulation scenarios using load balancing algorithm and without load
balancing. Four different attributes have been compared in both types of scenarios such aswith
load balancing algorithm and without load balancing.Considered attributesfor observation are
load distribution on each of the available network, average packet end to end delay, average
packet jitter values and handover latencies. The results showed that with load balancing all
three parameters showed improvement in the target heterogeneous wireless network
architecture; apart from average end to end delay, which increased slightly for the scenarios
using load balancing algorithm. As with load balancing not all the mobile users gets the best
available network therefore end-to-end delay for those mobile users go high degrading the
overall performance to some extent. Load balancing algorithm assures the fair load distribution
between the overlapping networks whereas without load balancing different networks show
abrupt load variationswhich decrease the performance with high congestion, high call dropping
probability and blocking probability at overloaded network. Load balancing approach utilizes
the available radio resources efficiently. Handover latencies are minimized, as it does not
require all the mobile users to handover when load balancing algorithm is used. Hence the load
aware RAT selection is a better approach as it offershigh radio links utilization with minimum
number of handovers and hence low handover delays, minimizedcall/connection blocking and
dropping probability and ability to maximize the network availability with uniformly
distribution of load in co-located networks. With load balancing algorithm there is an end-to-
end delay tradeoff, which is very small and in some cases negligible if underneath running
application is tolerant to very little increase in delays.
REFERENCES
[1] IEEE Std. 802.21, Media Independent Handover Services, 2009.
[2] N. Efthymiou, Y. F. Hu, R. E. Sheriff, “Performance of Intersegment Handover Protocols in an
Integrated Space/Terrestrial-UMTS Environment”, IEEE TRANSACTIONS ON VEHICULAR
TECHNOLOGY, VOL. 47, NO. 4, NOVEMBER 1998.
[3] A. H. Aghvami, P. Smyth, “Forward or backward handover for W-CDMA”, 3G Mobile
Communication Technologies, First International Conference, 2002.
[4] C. Chrysostomou, A. Pitsillides, F. Pavlidou, “A Survey of Wireless ATM Handover Issues”,
International Symposium of 3G Infrastructure and Services, 2001.
[5] IEEE std. 802.11, “Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY)Specifications”, 1999.
[6] Smith, Clint, “3G wireless with WiMAX and Wi-Fi: 802.16 and 802.11”, McGraw-Hill, 2005.
[7] IEEE std. 802.16, “Air Interface for Broadband Wireless Access Systems”, 2009.
[8] 3GPP TS23.246, “Architecture and functional description (Release 6)”, v.6.0.0, October 2003.
[9] 3GPP TS 24.007, “3rd Generation Partnership Project; Technical Specification Group Core
8/3/2019 Load-Aware Radio Access Selection in Heterogeneous Terrestrial Wireless Networks
http://slidepdf.com/reader/full/load-aware-radio-access-selection-in-heterogeneous-terrestrial-wireless-networks 20/20
International Journal of Computer Networks & Communications (IJCNC) Vol.3, No.6, November 2011
114
Network and Terminals; Mobile radio interface signalling layer 3; General aspects”, Release 8,
V8.2.0 (2009-06).
[10] Walke, Bernhard, “UMTS: the fundamentals / B. Walke, P. Seidenberg, M.P. Althoff”, John
Wiley, 2003.
[11] Kreher, Ralf, “UMTS signaling: UMTS interfaces, protocols, message flows and proceduresanalyzed and explained”, John Wiley & Sons, 2005.
[12] Shiann-Tsong Sheu and Chih-Chiang Wu, “Dynamic Load Balance Algorithm (DLBA) forIEEE
802.11 Wireless LAN”, Tamkang Journal of Science and Engineering, vol. 2, No. 1 pp. 45-52
(1999).
[13] IoannisPapanikos and Michael Logothetis, “A Study on Dynamic Load Balance for IEEE 802.11b
Wireless LAN”, 8th International Conference on Advances in Communications and Control,
COMCON'01, June 2001.
[14] Ted Scully, Kenneth Brown,“Wireless LAN load balancing with genetic algorithms”,
KnowledgeBased Systems, (2009).
[15] Mun-Suk Kim and SuKyoung Lee, “Load balancing and its performance evaluation for layer 3and
IEEE 802.21 frameworks in PMIPv6-basedwireless networks”, WIRELESSCOMMUNICATIONS AND MOBILE COMPUTING, Wirel. Commun. Mob. Comput. (2009),
DOI: 10.1002/wcm.832.
[16] SuKyoung Lee, KotikalapudiSriram, Kyungsoo Kim, Yoon Hyuk Kim, and Nada Golmie,
“Vertical Handoff Decision Algorithms for Providing Optimized Performance in Heterogeneous
Wireless Networks”, IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, JANUARY
2009.
[17] Alexandros Kaloxylos, IoannisModeas, FotosGeorgiadis and Nikos Passas, “Network Selection
Algorithm for Heterogeneous Wireless Networks: from Design to Implementation”, Network
Protocols and Algorithms, ISSN 1943-3581, 2009, Vol. 1, No. 2.
[18] Kamil H. Suleiman, H. Anthony Chan, Mqhele E. Dlodlo, “Load Balancing in the Call Admission
Control ofHeterogeneous Wireless Networks”, IWCMC’06, July 3–6, 2006, Vancouver, British
Columbia, Canada.
[19] Abdallah AL Sabbagh, “A Markov Chain Model for Load-Balancing Based and Service Based
RAT Selection Algorithms in Heterogeneous Networks”,World Academy of Science, Engineering
and Technology 73 2011.
[20] www.isi.edu/nsnam/ns/index.html.